Face Generation¶
In this project, you'll use generative adversarial networks to generate new images of faces.
Get the Data¶
You'll be using two datasets in this project:
- MNIST
- CelebA
Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.
If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".
NOTE: This solution draws heavily from DCGAN example from class materials available at
https://github.com/udacity/deep-learning/blob/master/dcgan-svhn/DCGAN.ipynb
#data_dir = './data'
# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
data_dir = '/input'
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper
helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)
show_n_images = 25
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')
CelebA¶
The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.
show_n_images = 25
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))
Preprocess the Data¶
Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.
The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).
Build the Neural Network¶
You'll build the components necessary to build a GANs by implementing the following functions below:
model_inputsdiscriminatorgeneratormodel_lossmodel_opttrain
Check the Version of TensorFlow and Access to GPU¶
This will check to make sure you have the correct version of TensorFlow and access to a GPU
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf
# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer. You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))
# Check for a GPU
if not tf.test.gpu_device_name():
warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))
Input¶
Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:
- Real input images placeholder with rank 4 using
image_width,image_height, andimage_channels. - Z input placeholder with rank 2 using
z_dim. - Learning rate placeholder with rank 0.
Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)
import problem_unittests as tests
def model_inputs(image_width, image_height, image_channels, z_dim):
"""
Create the model inputs
:param image_width: The input image width
:param image_height: The input image height
:param image_channels: The number of image channels
:param z_dim: The dimension of Z
:return: Tuple of (tensor of real input images, tensor of z data, learning rate)
"""
# TODO: Implement Function
inputs_real = tf.placeholder(tf.float32, (None, image_width, image_height, image_channels), name='input_real')
inputs_z = tf.placeholder(tf.float32, (None, z_dim), name='input_z')
learning_rate = tf.placeholder(tf.float32, name='learning_rate')
return inputs_real, inputs_z, learning_rate
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)
Discriminator¶
Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the discriminator, tensor logits of the discriminator).
Dimensionality Help (from class)
From what we've learned so far, how can we calculate the number of neurons of each layer in our CNN?
Given:
our input layer has a width of W and a height of H
our convolutional layer has a filter size F
we have a stride of S
a padding of P
and the number of filters K,
the following formula gives us the width of the next layer: W_out = (W−F+2P)/S+1.
The output height would be H_out = (H-F+2P)/S + 1.
And the output depth would be equal to the number of filters D_out = K.
The output volume would be W_out H_out D_out.
Knowing the dimensionality of each additional layer helps us understand how large our model is and how our decisions around filter size and stride affect the size of our network.
def discriminator(images, reuse=False):
"""
Create the discriminator network
:param images: Tensor of input image(s)
:param reuse: Boolean if the weights should be reused
:return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
"""
# TODO: Implement Function
#return None, None
alpha = 0.2
with tf.variable_scope('discriminator', reuse=reuse):
'''
Example:
# Input layer is 32x32x3
x1 = tf.layers.conv2d(x, 64, 5, strides=2, padding='same')
relu1 = tf.maximum(alpha * x1, x1)
# 16x16x64
Calculation:
W=32, H=32, F=5, S=2, K=64, P=?... we don't know P, would need to calculate....
# from tensorflow https://www.tensorflow.org/api_guides/python/nn#Convolution
out_height = ceil(float(in_height) / float(strides[1])) = ceil(32/2)=16
out_width = ceil(float(in_width) / float(strides[2])) = ceil(32/2)=16
'''
'''
input from reviewer
Use weight initialization: Xavier initialization is recommended as it helps model converge faster.
A possible implementation in tensorflow is to pass tf.contrib.layers.xavier_initializer() as the
value for the kernel_initializer parameter in tf.layers.conv2d
'''
# Input layer is 28x28x3
x1 = tf.layers.conv2d(images, 64, 5, strides=2, padding='same',\
kernel_initializer=tf.contrib.layers.xavier_initializer())
relu1 = tf.maximum(alpha * x1, x1)
# 14x14x64
x2 = tf.layers.conv2d(relu1, 128, 5, strides=2, padding='same',\
kernel_initializer=tf.contrib.layers.xavier_initializer())
bn2 = tf.layers.batch_normalization(x2, training=True)
relu2 = tf.maximum(alpha * bn2, bn2)
# 7x7x128
x3 = tf.layers.conv2d(relu2, 256, 5, strides=2, padding='same',\
kernel_initializer=tf.contrib.layers.xavier_initializer())
bn3 = tf.layers.batch_normalization(x3, training=True)
relu3 = tf.maximum(alpha * bn3, bn3)
# 4x4x256
# Flatten it
flat = tf.reshape(relu3, (-1, 4*4*256))
logits = tf.layers.dense(flat, 1)
out = tf.sigmoid(logits)
'''
Input from reviewer
Use Dropouts in discriminator so as to make it more robust. A possible implementation in
tensorflow can be achieved by simply passing the outputs from the last layer into the
tf.nn.dropout with a high keep_probability.
'''
return tf.nn.dropout(out, 0.9), tf.nn.dropout(logits, 0.9)
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)
Generator¶
Implement generator to generate an image using z. This function should be able to reuse the variables in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.
def generator(z, out_channel_dim, is_train=True):
"""
Create the generator network
:param z: Input z
:param out_channel_dim: The number of channels in the output image
:param is_train: Boolean if generator is being used for training
:return: The tensor output of the generator
"""
# TODO: Implement Function
#return None
#print(out_channel_dim)
alpha = 0.2
with tf.variable_scope('generator', reuse=(not is_train)):
# First fully connected layer
x1 = tf.layers.dense(z, 2*2*512)
# Reshape it to start the convolutional stack
x1 = tf.reshape(x1, (-1, 2, 2, 512))
x1 = tf.layers.batch_normalization(x1, training=is_train)
x1 = tf.maximum(alpha * x1, x1)
#print(x1.get_shape())
# 2x2x512 now
'''
input from reviewer
Use weight initialization: Xavier initialization is recommended as it helps model converge faster.
A possible implementation in tensorflow is to pass tf.contrib.layers.xavier_initializer() as the
value for the kernel_initializer parameter in tf.layers.conv2d
'''
# from quora:
# https://www.quora.com/How-do-you-calculate-the-output-dimensions-of-a-deconvolution-network-layer
# So=stride(Si−1)+Sf−2∗pad
# "where So means output size, Si input size, Sf the filter size.
# It is just the ‘opposite’ operation of the convolution (basically exchange the forward and backward pass)"
# So = 2(2-1)+5*1 = 7
x2 = tf.layers.conv2d_transpose(x1, 256, 5, strides=2, padding='valid',\
kernel_initializer=tf.contrib.layers.xavier_initializer())
x2 = tf.layers.batch_normalization(x2, training=is_train)
x2 = tf.maximum(alpha * x2, x2)
#print(x2.get_shape())
# 7x7x256 now
x3 = tf.layers.conv2d_transpose(x2, 128, 5, strides=2, padding='same',\
kernel_initializer=tf.contrib.layers.xavier_initializer())
x3 = tf.layers.batch_normalization(x3, training=is_train)
x3 = tf.maximum(alpha * x3, x3)
#print(x3.get_shape())
# 14x14x128 now
# Output layer
#logits = tf.layers.conv2d_transpose(x3, out_channel_dim, 5, strides=2, padding='same')
logits = tf.layers.conv2d_transpose(x3, out_channel_dim, 5, strides=2, padding='same',\
kernel_initializer=tf.contrib.layers.xavier_initializer())
out = tf.tanh(logits)
# 28x28xout_channel_dim now
'''
Input from reviewer
Similar to the discriminator, use Dropouts in the generator at both train and test time with
keep_probability as 0.5 as suggested here.
- this didn't work so well... reverting back to no dropout...
return tf.nn.dropout(out, 0.5)
'''
return out
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)
Loss¶
Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:
discriminator(images, reuse=False)generator(z, out_channel_dim, is_train=True)
def model_loss(input_real, input_z, out_channel_dim):
"""
Get the loss for the discriminator and generator
:param input_real: Images from the real dataset
:param input_z: Z input
:param out_channel_dim: The number of channels in the output image
:return: A tuple of (discriminator loss, generator loss)
"""
# TODO: Implement Function
#return None, None
alpha=0.2
g_model = generator(input_z, out_channel_dim)
d_model_real, d_logits_real = discriminator(input_real)
d_model_fake, d_logits_fake = discriminator(g_model, reuse=True)
'''
Feedback from reviewer
I would recommend you to multiply labels (for d_loss_real) by a smoothing factor (0.9, for instance).
This helps optimizing this loss for the following reason: initially the generator network does not produce
anything close to the real input images; hence, the discriminator quickly learns to distinguish between real
inputs and generated inputs - outputting a probability close to 1; hence cross-entropy loss will involve the
following computation: log(some_very_small_number), which can be unstable.
'''
d_loss_real = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_real, labels=tf.ones_like(d_model_real)*0.9))
d_loss_fake = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.zeros_like(d_model_fake)))
g_loss = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.ones_like(d_model_fake)))
d_loss = d_loss_real + d_loss_fake
return d_loss, g_loss
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)
Optimization¶
Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).
def model_opt(d_loss, g_loss, learning_rate, beta1):
"""
Get optimization operations
:param d_loss: Discriminator loss Tensor
:param g_loss: Generator loss Tensor
:param learning_rate: Learning Rate Placeholder
:param beta1: The exponential decay rate for the 1st moment in the optimizer
:return: A tuple of (discriminator training operation, generator training operation)
"""
# TODO: Implement Function
#return None, None
# Get weights and bias to update
t_vars = tf.trainable_variables()
d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
g_vars = [var for var in t_vars if var.name.startswith('generator')]
# Optimize
with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)
return d_train_opt, g_train_opt
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np
def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
"""
Show example output for the generator
:param sess: TensorFlow session
:param n_images: Number of Images to display
:param input_z: Input Z Tensor
:param out_channel_dim: The number of channels in the output image
:param image_mode: The mode to use for images ("RGB" or "L")
"""
cmap = None if image_mode == 'RGB' else 'gray'
z_dim = input_z.get_shape().as_list()[-1]
example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])
samples = sess.run(
generator(input_z, out_channel_dim, False),
feed_dict={input_z: example_z})
images_grid = helper.images_square_grid(samples, image_mode)
pyplot.imshow(images_grid, cmap=cmap)
pyplot.show()
Train¶
Implement train to build and train the GANs. Use the following functions you implemented:
model_inputs(image_width, image_height, image_channels, z_dim)model_loss(input_real, input_z, out_channel_dim)model_opt(d_loss, g_loss, learning_rate, beta1)
Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.
def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
"""
Train the GAN
:param epoch_count: Number of epochs
:param batch_size: Batch Size
:param z_dim: Z dimension
:param learning_rate: Learning Rate
:param beta1: The exponential decay rate for the 1st moment in the optimizer
:param get_batches: Function to get batches
:param data_shape: Shape of the data
:param data_image_mode: The image mode to use for images ("RGB" or "L")
"""
# TODO: Build Model
# def __init__(self, real_size, z_size, learning_rate, alpha=0.2, beta1=0.5):
exploring = False
if exploring:
print(batch_size)
print(z_dim)
print(get_batches)
print(data_shape)
print(data_image_mode)
#tf.reset_default_graph()
if exploring: print('here0')
#def model_inputs(image_width, image_height, image_channels, z_dim): return inputs_real, inputs_z, learning_rate
input_real, input_z, l_rate = model_inputs(data_shape[1], data_shape[2], data_shape[3], z_dim)
if exploring: print('here1')
#def model_loss(input_real, input_z, out_channel_dim): return d_loss, g_loss
d_loss, g_loss = model_loss(input_real, input_z, data_shape[3])
if exploring: print('here2')
#def model_opt(d_loss, g_loss, learning_rate, beta1): return d_train_opt, g_train_opt
d_opt, g_opt = model_opt(d_loss, g_loss, l_rate, beta1)
steps = 0
print_every = 10
show_every = 100
if exploring: print('here3')
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
for epoch_i in range(epoch_count):
for batch_images in get_batches(batch_size):
# TODO: Train Model
steps += 1
# Sample random noise for G
batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))
'''
Feedback from review
In the notebook there is a description of the "faces" dataset - it says that the values in the
matrix lie in [-0.5, 0.5] range. But your generator produces output in the range [-1, 1]
(due to application of tf.tanh). Hence, you need to multiply batch_images by 2 to achieve the same
scale. This fix will DRAMATICALLY improve your model's performance.
'''
batch_z = batch_z * 2
# Run optimizers
#_ = sess.run(net.d_opt, feed_dict={net.input_real: x, net.input_z: batch_z})
#_ = sess.run(net.g_opt, feed_dict={net.input_z: batch_z, net.input_real: x})
_ = sess.run(d_opt, feed_dict={input_real: batch_images, input_z: batch_z, l_rate: learning_rate})
_ = sess.run(g_opt, feed_dict={input_real: batch_images, input_z: batch_z, l_rate: learning_rate})
if exploring: print(steps)
if steps % print_every == 0:
# At the end of each epoch, get the losses and print them out
train_loss_d = d_loss.eval({input_z: batch_z, input_real: batch_images})
train_loss_g = g_loss.eval({input_z: batch_z, input_real: batch_images})
print("Epoch {}/{}...".format(epoch_i+1, epoch_count),
"Discriminator Loss: {:.4f}...".format(train_loss_d),
"Generator Loss: {:.4f}".format(train_loss_g))
if steps % show_every == 0:
show_generator_output(sess, 25, input_z, data_shape[3], data_image_mode)
MNIST¶
Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.
batch_size = 32
z_dim = 100
learning_rate = 0.0001
beta1 = 0.5
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2
mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
mnist_dataset.shape, mnist_dataset.image_mode)
CelebA¶
Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.
batch_size = 32
z_dim = 100
learning_rate = 0.0001
beta1 = 0.5
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1
celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
celeba_dataset.shape, celeba_dataset.image_mode)
Submitting This Project¶
When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.